Transcript: DjangoCon US 2025 Recap

If you're listening in the just the podcast, we're now doing video.

We're on YouTube.

Will's got this amazing self-back light.

You must check out.

It's just gorgeous.

I had to get glasses for the first time in my life two weeks ago.

And it's so

Is that is is that what they call?

Yeah, well it's two things.

Good for you.

It's like you're married to a doctor or something.

So yes.

So uh

So um it means elder.

The press the Presber bit means elder.

So there we go.

Well now we've just lost half our audience, but that's okay.

Django Web Framework.

We're back, trying new things, and you know, six years in, almost seven years in, we're just gonna mix it up.

We're gonna do videos, we're gonna do guess, not guess.

Um yeah, just have fun with it.

So the big news, yeah.

We'll go back to our regular cadence of uh twice a week or every two weeks, fortnightly, as they say.

There's a perfect word for it.

If only it was internet.

Like you sound pretty pretentious to say it to American.

You could say it, but I if I when I say it, it sounds

Yeah.

Um all right, so let's dive in.

So we're gonna talk about, we're back.

Uh DjangoCon US just happened last week, which I was at and I was live texting you the whole time.

Yeah so there's some things to talk about, APIs maybe coming into Django and other things.

And we're also we're gonna do our ads.

So we we're very thankful to have a sponsor, Hacksoft.

Uh, you're a Django development partner beyond code.

We're gonna talk about them a little bit later.

Both of us had a chance to spend some time with them at DjangoCon Europe in April.

Yes.

So, Carlton, summer.

Go.

What was what was the summer like?

We haven't, you know, without the podcast, we don't we don't talk like we used to.

You know, this is just kind of an excuse for us to you know, texting only gets so far.

It's just uh it just doesn't work.

Like the summer here is long because the kids finish school the last week of June and they don't go back till like you know the second week of September, and so that it's uh it's a long break.

Um, but I really enjoyed it.

Um we were busy over the summer.

Um I can spoil it.

We'd had Google Summer Co.

coming on.

We were getting ready for you know, helping to um get some of the features in there for Django 6.0, which we'll talk about in a minute.

But it was just a nice relax, it was quite busy.

I had to take August, like July was very busy.

I had to take August a little bit more on the back burner because I just needed some rest so that I could get I could arrive now in September and actually be fresh.

So you feel fresh?

I mean, I went I was away, I had a week's holiday with my eldest in um first week of September.

He wants to go to university in France.

So we did a lightning tour of French cities.

It was wonderful.

I got back and I was quite tired.

I mean, you know.

Yeah.

I mean, and it's it works.

I mean, it's not it's not the quick quickest way of learning a language, I suspect.

And it may not be the best way of learning the language, but I could have been playing Candy Crush in that time and I wasn't.

Yeah.

Um well, that's good.

It's nice to see you back.

I'm just staring at my notes uh trying to remember, you know.

I I had to think about it.

So some people at DjangoCon US told me there there's an updated version of bingo for this podcast where I have to we have to mention Jeff Triplett, and I have to say I worked at I work at uh PyCharm.

So I'll get that out of the way.

So I got to go to EuroPython in Prague for the first time.

Beautiful city.

I can't say I saw a ton of the city, but it's a work trip.

They're all teenagers now, and they move on to their next level and they're less slightly more independent.

I can become kind of old man of the European conference scene.

I've never been to EuroPython, but it always looks amazing.

No, it was great.

It's just like the Yeah.

No, it was beautiful.

And Chicago is actually in a way similar to that because there's a big river near where my hotel and the venue was, except Chicago is like 50, 100 story buildings in Prague, a little bit more manageable.

But you can walk along and there's cafes and it's really, really nice.

Um and Chicago, yeah, in September is

Oh, yeah, yeah.

Yeah, yeah.

I was I was running by that with uh with a colleague in the morning and we're like, what is that?

You know, because it looks it's over it looks over a hundred years old, and now they have these beautiful cafes and bars in it.

Um I was like, they surely didn't make it, you know.

And these these

uh if

So yeah, amazing city.

Um, excuse me, really good food and everything.

I have this new setup trying to get going.

Um, what else?

Um you were invited, but you were busy.

There was a Django 20th episode on Talk Python.

So um Adrian Halavati, Simon Willison, Jeff uh Triplett, Thibaut Colas, our president, and me, even though I was like, don't put me on here.

Like, but you couldn't attend Carlton, so I got your spot.

So that was really nice to sort of go back memory lane a little bit between Simon and Adrian and Jeff, um, to be fair, since they were all have been there from the beginning.

I think we should link that in the show notes.

It's a nice, uh it was a nice chat.

Um, so if you didn't catch it, like I'd I'd recommend it's worth the you know half hour and two X certainly.

Yeah.

Uh but yeah, no, I think a lot of Django today, you can see in, you know, Adrian talked about this, the fact that it was built for professional use case, and his and Simon and Jacob's design philosophies and just personalities, you know, are really clear uh in the framework today.

So Jacob Kaplan Moss was at DjangoCon US, which was great, and he's on the board and very involved.

What else?

UV live stream with Michael Kennedy.

So that was through PyCharm.

That was fun.

Uh a little nerve-wracking to talk for an hour or talk for an hour, but I was the one.

Yeah.

Yeah.

Um, Django survey is coming up.

That'll be released soon.

We're gonna do a show about that.

I think the next show we'll do on that.

So they'll the results are ready.

I gotta do a blog post.

Hopefully, I took some videos at DjangoCon US.

I mean, so yeah, I I wanted, you know, JetBrains mentioned Jetbrains, but I really think jet brains do a wonderful thing.

I like they they put an awful lot of effort into it.

It's a big thing, and it's every single year that the Python survey, Django survey.

I just, you know, little round of applause for the effort and commitment that jet brains has shown for that over the years, I think.

So that's that's historically been a you know a quarter of the annual budget.

So that's yeah, it's a big thing.

Um and then lastly, my personal site, I've been trying to write a little bit.

Uh so wsvincent.com and it's it's a Ruby Jekyll site.

That's how you know it's old.

So I just I've been trying to vibe code it just because I I wanted something that matters, but I can sit back and just be like, all right, agents, like do your thing.

So yeah, it's got new search and

I'm not gonna jump on it.

like wet, like it is every

I just can't be bothered and use Django.

I know, I know.

I will.

That's doing great despite my neglect.

So that's yeah, that's going up and to the right.

It does great despite your neglect.

So anyway, so yeah, busy summer, and let's talk about DjangoCon US.

So that was last week and I was there the whole time.

Maybe I'll I'll tee it up and then you can or I don't know.

We can discuss it.

I know.

I well I missed I missed you at the conference.

It's not the same.

And normally we play that game of someone comes up and says something nice about the podcast or something, and we can kind of you know hand them off to each other as a next step.

Because that was I would be like, Carlton, I talked to so and so, so and so-and-so.

You're like, oh, say hi.

Sometimes I showed the I showed the phone, I showed the you know the I actually had said hi.

Yeah, yeah, yeah, yeah, yeah, yeah.

Um so Carson Gross, uh creator of HDMX, gave the keynote.

And this was uh did I link?

Yeah, I wrote actually I I just wrote up uh a written review of the conference, um, which we'll link to where I mean it was a great talk.

It was about this Japanese um uh designer at who worked at Nintendo and his concept of withered technology, which is sort of an older version of boring tech.

And Carson made a lot of great analogies between HTMX, hypermedia, and you know, just with experience, you don't have to you can focus on sort of delight rather than just technical brilliance for the sake of technical brilliance.

Like to miss the conference, to miss the community, yes, those things.

But then to miss Carson, who um with HGMX is a bit I've been building a product last two and a half years, I've been using HTMX since day one, and I am just so on board with it.

And it's for me, it's so right, and it so speaks to what the world the web is and can be, and it fits so well with Django.

Yeah.

And to not be able to, you know, and I ban to with Carson Online and to not be able to meet him is like, ah, I can't believe it.

Well, he's I mean, he's a he's a Java guy uh and JavaScript.

So he's he's a professor in Montana, so he doesn't do much Django himself, but you know, this HTMX is there and the Django community has adopted it more than almost more than anywhere else.

Um so yeah, it was good.

I mean it's it's unfair to say it's the new jQuery, but it's a little bit, you know, it's just like sprinkling your inner activity.

And he also

sorry go ahead.

Um was about uh you know about my experience of building with um both HTMX and Alpine the last couple of years.

Um one of the lines that one of the points I make is I haven't had this feeling since the old days of using jQuery.

And it was wonderful.

It was this glorious world with Alpine and HTMX.

I've really found that again and it's just joyful.

I mean, it's just whatever you need.

I'm just looking at my time.

And I'm interested by the people to talk to the people who use those different options as to what they what they're getting, which HTMX doesn't offer.

And I think I think maybe they're slightly higher level, they perhaps have better encapsulations of some things that you have to do manually with HTMX.

I'm not sure because I haven't really dug into

We I had some great hallway chats with several people talking about you know the combination of HTMX and Alpine, and there's a couple edges there, like if you want to go automatic back refill.

I don't know, I took some notes on it, but he has I mean he's a smart guy.

He's got less to say.

But if you've got an Alpine component and inside it there's a little bit of HDMX that you swap out and put into fresh bit, Alpine can get confused.

You have to you have to sort of either use the morph plugin so that Alpine knows, or you have to go back to Alpine and say, hey, by the way, isn't it this bit of tree here?

There was there was some tip about with a back button, and he was saying you can just turn history off.

Um so he's got lots of tips.

He was I was I was sitting in a group of people who've been using it much more seriously than I have.

Uh so anyways, we'll try to get him on.

But yeah, great talk, I think aligns with the ethos of Django.

And no, and I I it's it I mentioned to you, I think it's worth saying.

So in Django Land, HTMX is the hotness.

Like all the top talks the last two years.

If you look on YouTube uh for Django Cons, HTMX is is the top one.

And yet someone walked out, I walked out with someone from that talk, and they they turned to me and said, I'd never heard of HTMX before.

So this is the echo chamber, right?

Is we're all online and we're doing this and we know to each other.

So this I'm trying to just get more comfortable with repeating myself.

You know, say it in the podcast, put it in the newsletter, put it in a thing, you know.

I mean, give people the chance to cut me off because I hate people telling me something I've already heard, but um, you know, meet people where they're at.

It is yeah.

Um, but now I wanted to ask you, because your talk, you gave a talk as well, you're right, uh AI talk, which we're going to, but you were run right after cast.

Yeah, yeah.

So there's an upstairs and a downstairs, which is unavoidable with the venue.

It was maybe the only thing that if I could wave a wand and change the, you know, going up and down the stairs meant it was a little divided.

So, but

Um it motivated me to put HTMX in my demo because I was like, ooh, this'll be so I was I was quite pleased.

I and when I present when I uh submitted the talk, I thought I would just do the a version of the DjangoCon Europe one I did, which was Django for data science.

But then I was like, ooh, Django for AI, because you know, kind of what's the difference?

But then that got me thinking, oh, I have to, yeah.

So it's two-thirds new.

So basically the whole DjangoCon Europe talk where I talked through a lot of the code.

I really didn't want to talk through views files, you know,

20, 20

So I it's sort of a two-parter on classic machine learning, uh, which was the DjangoCon Europe one, talking about how quickly you can train a model.

I use the iris data set, get a job lib file, and then Django's perfect to deploy that, throw some forms, store the data in a uh database.

I think more data science people should know that Django's there for them.

You could you could clone the repo and vibe code almost any model in there.

And I think that

would so that would help.

It's it's something I can't get hard numbers on this because they're all private, but it's at least 10 times more computationally expensive to do to talk to a chatbot than it is to do a Google search.

And right now they don't monetize, you know, they only monetize around 2%, you know.

So I I resisted going into an economics kind of thing on it, but it's wild.

I mean, the amount of build-out both to train, but even more so to serve these LLMs is is unprecedented for sure.

Um massive contrast between the revenue, the capex expenditures, and then the revenues.

And it just seems that the revenues are, you know, maybe they've got a dream of how it will come, but it just as a sort of, you know, looking at it from afar, I'm like, ooh, that stinks.

It wasn't

like a

What's the third thing?

Um Data.

There's a third one.

Everything comes in.

So parameters like knobs you can turn.

Um and so that's what that's like the B at the billions of parameters, and it's probably trillions for some for GPT-5 and stuff.

But yeah, I don't know, it's all interesting.

I mean it's different.

I think for me, I learned a lot researching the talk, and I think most web people aren't AI developers, so it's good for us to have a sense.

I mean, one of the interesting things about parameters is that's like you s when you start, you you you pick a number of parameters that things you can turn.

They get you know, more and more generally is better, but you can't just put a quadrillion because if you have too many and it you don't have enough data, then it just repeats verbatim the training data.

So that's why you need more data.

Yeah.

Well, and I mean GPT-2.

Now I'm getting on a rant, but GPT-2 was was trained on Reddit, on Reddit stuff, just like almost on a laptop.

Um was uh Reddit post with three, three votes or more.

And so you focused on the inputs and then the outputs took care of themselves.

Now it's just vacuum up everything.

And so it's a mess, right?

And so what they're doing is they're adding a filter on top of the model to make it available for public consumption.

But to train that filter, they're using um human-led reinforcement learning.

I'm getting that wrong, but they're getting humans in the global south to like review horrific stuff to put a filter on top because the inputs are kind of garbage.

So it's, you know, it's all the old colonialism stuff.

But anyways,

So then you can't.

But but the cool part I thought I thought with the demo was I I I wanted to uh I'm quite keen on local models.

So OLAMA, you can pull in whatever local model you want.

And I rigged up a demo, there's a GitHub repo, where you can pull in whatever model you want and then use server sent events to have a chatbot that streams the the tokens, which is the key part I wanted.

Um and yeah, just to show Django as is can do a lot.

Like you don't need async views for it.

You don't um, you know, I also played around with HTML streaming, which is kind of cool.

You would probably use an async view.

And but it more importantly, I used HTMX, so I had like an interactive demo uh which I thought was cool.

Yeah, yeah, yeah, yeah.

Exactly.

I just wanted to, I thought the your I read up your write up of your talk.

I'm interested to watch the the final thing.

I think the demo is super cool.

Um, but the you had this um the preliminary chat where you're kind of placing Django again in a wider context, and you had this nice slide about how um the web is not the same as API endpoints.

And yeah, I think that's like that's really the drum we should be banging, right?

Yes, okay, you spin up an API endpoint, you connect it to your IM, but a web application has so much more going on for it.

And that's our strength.

But yeah, you need everything.

And you can kind of do it with fast API.

There's there's

you know authentication.

There's actually a EuroPython talk that I think just came out where someone talked about combining all of them.

So yeah, so I think it's a real opportunity for Django.

I think, and I I'm trying to learn more about it and think about how we can meet them where they are, because one of the other big points was this new generation of uh young Python AI developers don't know the

way and they and they think

But you know, to Carl to Carson's talk, like Withered Tech, like Django's here for 20 years.

Like we've got a solution, maybe not the best solution, but we've got a solution to a lot of these problems.

So I want to

We're not gonna be running them ourselves.

But what does that look like?

You know, sometimes it's streaming, a lot of times it's not.

It's just medical records, right?

Just send it over, get a response.

It's feels like it could just be kind of standard REST API stuff.

Um you s you were saying that Carson Carson was talking about young developers not even knowing what HTML forms were.

Well, I I felt I felt a kinship, not just because I think we're around the same age, because I was trying to bang some of those drums.

Um and yeah, and I think and he said, and I agree that the onus is on us with gray hair, you know, I got I got a little bit to uh you know to try to meet people where to try to meet people where they're at, you know.

It's not um because again, I also going to these Python conferences, people would I had another slide about like I was like, well, why not Django for your web needs?

And they're like, Well, Django's slow, Django's old, and it's hard to learn.

And it's hard to know where to start.

Yeah, you know.

No, no, for sure.

Yeah, no, and maybe some books that you know have examples of of this in their next editions.

Yeah.

I think, you know, it's great to go to these conferences, but we also have to remember it's 300 people out of probably three million or something that use Django.

And, you know, just because we all know what HTMX is and we all know these things, we need to speak outward a bit more.

So hopefully this, you know, the video version of our podcast, and maybe I'll do some videos to

reach people where they are.

Wait, and do you want to do the the mid-roll, Hacksoft, who's sponsoring?

Do you want to read out?

Hacksoft is your development partner beyond code.

From custom software development to consulting, team augmentation, or opening an office in Bulgaria.

They're ready to take your Django project to the next level.

Yes.

Also the Django newsletter, in general, just good citizens of the Django community.

Yeah, and

They've been around for years.

So beyond the sponsorship, I think Hacksoft is a great team.

Yeah.

Yep.

All right.

Now now we're gonna let you talk.

So Google Summer of Code.

So uh Carson mentioned he was really excited about template partials.

Yep.

Uh and

So, you know, um how can I say like templates are so I sat down to start the new project back in the day, like 4.2.2 was around so um early part of um 2023 when I stepped down from the fellow role.

And I'd read the essay on the HTMX website about template fragments, which I call partials for you know slightly different reasons.

Um, but they go by this fragments or partials name.

And I I was like, we need that.

And there was a couple of options in the um in the ecosystem.

There's one called um Django render block, I think.

Um J I can't remember exactly what it was called, but it had a render block to string function, which you would call a block, you know, you'd say you get a named block from your template, and you you you get it, and you'd render that and you could return that, but you had to adjust the the flute, the the view flow.

You had to uh instead of returning a template response with just the template name and the context, and then you know, whatever magic Django does to actually render that behind the scenes for you, you would um render the con the template and then pass it to an HTTP response and you return that.

And it's quite disruptive to your few logic, like the generic views, you you're basically rewriting every one.

Um and that wasn't that wasn't what I wanted.

Um and so I I built template Django template partials to um a slightly different take on that, which is how I wanted it, and integrated that with the template loader, and I sort of announced it at um Django Con Europe in 2023 in Edinburgh.

Um and it was like you know, the response was phenomenal.

And so you know, a whole LTS cycle later, we've got um Fahhan who's come in and done the Google Summer of Code project to merge it into Django Core, and over the summer he was working hard and some nice tricky issue.

Temple partials have been working fine for two years, no, you know, no issues at all.

We had I first released it, we had a bit of feedback, changed the IPI, etc.

It's been working perfectly for everybody.

But that step up into getting into Django, it's like, oh, there's quite a lot of resistance there, and it enough resistance such that I'd kind of cracked open a branch and started off at it a couple of times, but then gone, oh I'm too busy, I haven't got time to do this.

And I think without the Google Summer Code project, it probably would have just sat there indefinitely.

But with the Google Summer Project, uh co-project, Fahhan has has brought it up to scratch, merged it in, and it will be part of Django 6.0, which is feature freeze tomorrow when or to tomorrow as we record this, today as we release it.

Um and it will be the alpha will be there, and then the final release will be in December.

So, you know, fellows don't have to just retreat to a cave, they can, you know, do what you've done, right?

You you said right after five years you had all these ideas because I remember you just like sort of launched it, and I said, Wow, like that's that's quite a thing.

And you were like, Well, I just did it in an afternoon, but you've been stewing on it for years.

Like the Neapolitan, which is the other project I released at about the same time.

I'd lit I'd literally been sat on that for eight years, like as a sort of thing that I would need.

And so when it came to when it came to, oh, do you know what I'm gonna have to end up write actually now back in back in business writing fresh generic views for lots of models?

No, I'm not doing this the old way.

I'm doing it.

I'm I'm putting Neopolitan in place and I'm doing it.

It allows you to define your model, define your your Crowd view to specify some fields, and then all of a sudden you've got the create view, the list view, the retrieve view, the update view, etc.

So it's just just a different of a approach.

But we're you know, we we we kind of do the same, you know, like uh David Hannemeyer Hansen gave a gave a keynote a couple weeks ago talking about um crud monkeys.

And I'm not sure if I love using the word monkeys, but I take his point about we're just kind of doing crud and there's nothing wrong with that.

And then there's just tweaks to it.

And so this this yeah speeds things up for those inclined.

And it's like nobody makes a six-figure salary just writing a CRUD app.

You get all you the you get the six-figure salary for the bits after you've got the CRUD app in place.

And what I I love about Neapolitan is I can bootstrap it in seconds, and I've got the basic CRUD structure in place.

And then as I'm experimenting, as I'm evolving the application, as I'm I've as I'm getting into the details of what makes it a valuable piece of software rather than something that could be automatic automatically generated, then I can customize and it's got all the right hooks and all the things.

And that's sort of

But yeah, it's what next is what matters.

Um CSP?

Well, let's talk about so Django 6.0.

There's three killer features.

Um I'll let you know.

So there's for years it's been this third-party package, uh Django CSP, which has enabled you to set this.

And CSP is like um it's a defense in depth measure.

So if you temp if you use um you know the temperate language correctly and you do all your auto-escaping, you don't turn that off, more or less safe from most cross-site scripting attacks.

Like it's very hard to smuggle one in.

But as a defense in depth measure, you CSP elabors you to say things like only load JavaScript from the same domain as the application comes, or only from the same domain as the application, and this one other specific domain that I've told you about.

So if by some accident you you've um unescaped Java Unescaped script tag does make it to the page, the browser will see the unescaped script tag.

It will say, no, this domain, this you know, evil.com isn't allowed, isn't included in the content security policy as a source for um scripts, so I'm not going to load it.

So it's kind of a defense in depth measure.

Um it's massively complicated.

It's like there's just so much to it.

It's not it's a battery is like, you know.

It's nothing you want to do on your own.

You do not want to be implementing CSP protection by yourself.

And Django CSP, the third-party package, well, has been maintained well for years, but there was a period when it wasn't.

And uh, so what's what's happening there?

And for it to be in core, I think it's you know, you talk about battery, I think it's one of the parred batteries that it's good that Django provides.

And so I think it's a major feature, and I'm quite excited to be using it.

I I'm I was I was joking on um Mastodon the other day about whether it's too early to deploy 6.0 before the feat before even the feature freeze, deploy it to my production environment.

And it's because I want to move to the um in-core CSP protection rather than the third-party package.

And I don't really want to have to wait till December to do that.

So I'm now thinking, okay, so can I can I release this early?

You don't you don't need permission.

So that's not a problem for me.

So that's the main one, I would say.

Then then the next big feature is I think that the we're gonna um merge these um the backend, the interface for Django Tasks into um Django itself.

So Django Tasks is this um was this proposal from Jake to merge um or to create um a background task queue for Django.

So you know, if you want to send it, the classic example is you want to send an email.

And you saying an email might take a few seconds by the time you connect to the server and do all those things.

So you don't want to do it in a request.

You want to take the request, put the task to the side and say, I'm gonna send that later, return the response quickly to the user, and then have a worker in the background that can send off the email or generate the PDF or whatever the background task is.

That's been in development for a cycle now.

Um and the the interface bit of it will be merged into Django in 6.0.

You'll still to actually use it as a background task worker, you'll still need to use a third-party package where the database back end.

So the default option will be to store tasks just in your database.

So you write um do this task to your database, and then the worker fetches it from the database and does it.

That's still a going to be a third-party package, you're gonna need to install that as an extra.

And that's just to give it time to mature and to make sure that it meets it, it's tested widely enough to meet Django's stability guarantees.

Um the last thing we need to do is merge it very quickly and then find that it's flawed in some way that causes problems to existing deployments.

Nice.

So though, yeah, those are

Sorry, Jake.

While you were talking.

Um Yeah, so very exciting.

Actually, and we should we should mention also I got a chance to meet Jacob Walls, our new fellow, who was at who was there.

Um you played a role in that, right?

This was why this was the main reason I think why my July was ludicrously busy, was that um we had over a hundred applicants, um, which is amazing.

And it, you know, of those hundred applicants, we had many which were very high quality.

And it just goes to show the strength of Django and where we're at.

And you know, sometimes we wring our hands, are we okay?

Are we doing no?

Jang the it's it was it was amazing to see the the the number of really high quality applications for the fellow role.

Um, and we had to narrow it down, and we had to do interviews, and we, you know, it it was quite a big process.

But Jacob's gonna be amazing.

I got a chance to spend some time with him there, and and also Natalia Bedar, uh second of the three, also Sarah Boyce was there as well, and she was during the sprints, she was sort of letting him take the lead and scurry around as I was seeing you do for years and years and years.

But no, he's yeah, I mean, he's a very, very, very smart person.

Um, so I think

So, one more big thing we want to mention.

Django on the med.

So there was talks on this, Paolo Malchiori mentioned this is I know you're excited about this.

I'm excited about this.

It's three days to get together and work on Django.

I mean, uh my wife said the tagline needs to be not as exciting as perhaps it sounds, um, because it's uh because it's geeks in a room.

She's English too, right?

Um what is it?

It's yeah, three days to get together and work on work on Django, basically.

We're gonna um start eight at eight o'clock till about two-ish.

Um nice breakfast when you turn out, coffee, some croissants, these kind of things.

Make sure that you're fired up for the day.

Um, three days, and then in the afternoon do a bit of exploring, a bit of socializing, a bit more talking about the ideas, you know, and evening meals and all the things like a conference, but without the conference bit.

Um there was a nice uh post on the Python blog from uh Lukac, who's the um the Python developer in residence, um, saying that uh you know the sprints are the best bit.

And I'm like, yeah, no, they really are, and that's why you should come to Django on the Meg, because it's just um so we're the first day we want to do a roadmap session.

Um so we get we're gonna have people, we're gonna hopefully we're gonna have a fellow, we're gonna have a couple of board members, we're gonna have a couple of steering council members, we've got we've got really the absolute A team of you know, some of my all-time stars are turning up.

Um, it's gonna be amazing.

And we hope to drive forward the roadmap, um, sketch out some of the ideas.

And it's not just like the it's not just like, you know, if I say to you, one of the top community priorities is modernize contrib.auth, right?

Okay, well that just sounds like an absolutely impossible project that we need to get Andrew Godwin and we need to lock him in a cave for six months and he'll emerge with the

Kickstarter, yeah.

Um, we need to have modernized contrib authors and we need to break it down and spend some time working on well, what does that mean?

What are the what are the sort of pillars of coming up with a serious plan for for um modernizing Contribut to authors say as an example?

Or, or you know, um adding typing to Django?

Do we want to add typing to Django?

Okay, if we do, what does that mean?

Uh what are the approachable routes into that?

What would an experiment look like?

What would uh all of these questions, which until they rem until they're cashed out a little bit, we we can't really make progress, which is why you know modernized contributors, it's sort of sat there on the on the back burner for a while, but it's not really made progress.

And if we can then the third step, so the first step is identify the core roadmap buttons.

The second step is to break them down into you know something a bit more addressable.

And then the third step is if you've got those identified tasks, then you can put out the call for volunteers.

And so for years and years we've had this problem where volunteers would turn up, they'd be like, I want to help.

And it'd be like, well, we haven't got anything to point you at.

But if we've if we've had a bit of time to flesh out some of the core roadmap ideas, the things that the community would like to work on, then we can say, look, do you know what?

There are half a dozen identified things which are addressable to a new contributor.

Yeah, and it's not that we don't have volunteers, it's that we don't have enough guidance for volunteers.

Yeah, yeah.

I've got I'm hoping to organize a special birthday party treat for Django, and then more time to just work out.

I've got I'm gonna get, you know, um person A and person B in a room, and I'm gonna lock them up together with loads of quasar coffee, and I'm gonna say you're not allowed out until this has happened.

Right.

So in where you are in Pescara in Italy where Palo is based.

So the first one just happens to be in Paraviget, then we're gonna do it next year in Pescara, then you know, maybe twice a year would be good.

Um we've got to think about how it clashes with um agenda for say DjangoCon Europe and things like that.

So we need to consider, but yes, that's the idea is to do it.

You know, one in Palo Fuge, one in Pescara, and you know, maybe we can go somewhere else if you know uh Cyprus.

I know some member of the Jang community lives in Malta.

We could have Django and the Med in Malta.

Look, the last time we did a roadmap workshop, um, Thibaut um Colas is DSF president, he organized one and it was a couple of years ago, and it's on the forum, you can search for it, perhaps we can put the link.

But the number two um impactful item that was that came out was to do active developer sprints for you know Django's active active developers, active contributors.

And with that's what we're doing.

We're putting together the active developer sprints.

And I want to say to people, if you're a new contributor, if you're like if you'd like to get involved, you're you're welcome too.

You don't have to be an established contributor.

I'm excited because some of my superheroes are coming to Django on the map.

But I'm also, I want you know, you to come if you're if you're like, do you know what?

I I want to get involved too, because the other goal is to get the new contributors to come and help them get on boarded and joined with the experienced contributors.

And so by the time we've left, those people are you know up and running as well.

Right.

And that's in October, 7th to the 9th.

in the show notes.

And yes, Active Developer Sprints was number two.

And and and the biggest thing is like, okay, let's if we can do it once, then we can do it, you know, time and time again.

And the exact details don't matter so much as is it possible, is it feasible?

And yes, it is, and we're doing it.

It's gonna be a good idea.

I think Paolo with Plone, I mean, other communities have and do do these things because they have the same problem of, you know, the sprints at the end is helpful, but it's not quite enough time, not everyone's there.

You know, three days, but also three days coming in with ideas as opposed to like the end of a conference, you know.

I I've seen it takes, you know, the Thursday is helping people get set up and people do stuff, and then Friday sometimes there's a great burst, but usually people are it's a little more solid.

Yeah, you can buy.

So um yeah, so it's not

And Panamijay is it's a pretty nice place, you know.

I know you've been you've been to my hood in Brookline Mass, which is pretty nice, but in a different way.

No, I mean I I I like that whole red brick thing.

And it was autumn when I came and the sun was shining, so it was awesome.

Wow.

But yes, I'm excited about Jango on the mid um.

All right, so we we just

Yeah, so he so actually I flew out, he's based in Boston.

Um, he gave the Tuesday keynote, and we flew out together and he sort of previewed it for me.

And I was um you know mentioning his one of his ideas was bring Jinkover's framework into core.

And I think I have a little scar tissue from being on the board around that, so I was a little bit like, yeah, we should do something about APIs.

I don't know if that's gonna work.

But he got a really great response, and there was a huge long forum post, um, and he had a bunch of ideas.

So I'll toss it over to you, Carlton.

So um the

general idea

And I know that second part is not quite what you would say.

Like, so the forum post was Django needs an API story or something like that.

Absolutely.

And his option three was document the the the options we've got, which is DRF and probably Django Ninja are the two.

Ninja, for sure.

And yes, we should document the options.

And yes, it shouldn't be the case that you discover Django and you can't find on the website easily, you know, what you should do if you just want to you know arrest endpoint.

How how do I do that?

That shouldn't be a mystery.

You shouldn't have the the the absolutely we should document all that.

But had we merged DRF into Django Core a decade ago or even five years ago, it might it might have been a good idea.

But now things have moved on.

Like DRF is dated, it's the last generation, and I I I understand people are using it, but it's like if we just merged that into core and said this is the Django solution, I think that would be us saying we give up on innovation, we we we we we quit.

I I I I think there are demonstrate, I mean Django Ninja shows that much more is possible.

And yet I don't think Django Ninja is the final solution.

I think there's a whole number of reasons why it's not perfect, but it it it was fresh and it it demonstrates just absolutely clearly um that the other frameworks out there, the fast APIs, the light stars, the oh

So if I if I want to serialize it, so what is what's the main problem with DRF?

The main problem with DRF is that the serializers are last generation, they're slow because basically they loop through the fields that they're given and they loop through the object and they do the get it to it.

Whereas if you you you um you take catters, what it does is the first time it comes through to your serialize, it kind of generates code which is as good as if you wrote it by hand.

It's as optimized.

So it's kind of native speed and it's just much, much, much, much faster than um the REST framework serializers that we we know and love.

And the Django forms, the same sort of ideas, but forms aren't generally used in performance critical ways, whereas people with APIs they're they really often are needing fast serialization because they're doing thousands of records you know over the wire as fast as the API can respond.

So it it kind of matters.

Um and so whether it's Pydantic, whether it's catters, whether it's message spec, there's there's the whole my point is that there is a whole collection of next generation solutions out there, which Django, if we you know, one of the goals of the DSF is to push the state of the art forward.

Well, we don't do that by merging a last generation solution into Django.

Yeah.

Right.

And so this is hopefully getting people in a room, right?

So because nobody wants to be a blocker on this stuff, but it it's you know, the implementation details matter, right?

Like people, there are developers out there who literally hate Django because of DRF.

Right?

That's another whole separate thing.

Yeah.

And look.

DRF is wonderful, and I spent a lot of time working on it.

Well, just

I mean, I I I pop my head in every so often if I'm needed, but basically, I've moved away from it now.

But I built an entire career around DRF before I was the Django fellow, and I built um web applications for you know some of the biggest companies in the world using DRF, and it was a wonderful thing, and I learned so much from it, but it is out of that, and there is no question to that.

There is no question of that

at all.

I mean, I my takeaway, like I was talking to a number of people.

There's there's a working group on the website.

You know, I think if I had a magic wand, the two things I would do is, well, executive director, which people agree with this.

Um, the board actually discussed, they've done so this this is one of the things the board has done a lot of work around this, and something fell through that they spent a lot of time on.

And that's something worth mentioning because when you're on a board, you know, they spend hours and hours and hours and hours, and then it's just for reasons they can discuss like nothing bad.

It didn't, it didn't happen.

And that can feel discouraging, but it's all important because it gets us close, like we need one, we'll find the right person, we'll find the timing, you know.

But that is the kind of stuff that I hope the board, and I think and they're doing a much better job than when I was on it, talk about the work that they do because it's thankless work, it's important work.

Um, so executive director, and that was raised numerous times.

Um, the second is I would, and I think I'm even gonna do this.

I had this, I think I told you the idea.

So the homepage, right?

Like we can't change everything.

I was like, what about a third-party package homepage, right?

Like I could, when I have a moment and I'm truly procrastinating, I'll I'll vibe code.

Um, I would like four things on the homepage.

Well, what what do I want?

I want to look modern.

We all agree it doesn't look modern.

You know, if you look at Media JS, you look at View.js, those look, they have colors, you know, more modern.

Color, color.

Yeah.

because you

Um, but then the four things we need is we need, you know, maybe even it's just ugly boxes somewhere.

We need API and just links to like an API story, CRM, which Django CMS, Wagtail, like we have fantastic story.

There's other ones, those are the two big ones.

Um I would like to see a hello world tutorial on there, you know, not just polls.

And then AI.

And it maybe it just links to a page where it talks about how you can use Django where it's appropriate, but just an AI story.

I think you can't be a web framework in 2025 without an AI story.

And I think it's important to

Like people think, oh, Django can't do async.

There was a um a post recently about um why isn't async.io more popular than it is, and like the sort of headline bit was oh, Django's async is underdeveloped.

It's like it's really not.

The only bit left is the is the cursors to the to the actual ORM, right?

The the rest of it all works.

So if you want to do wiring up to an LLM, async views, they're wonderful for you and gonna work perfectly for you.

But that's not the narrative that's out there in the ecosystem.

People are like, oh no, Django's a uh async.

It it is sort of interesting that I think that was always the issue with well, WebSockets with two-way and async was a great technology, but like how common was the use case?

Yeah, yeah.

And you know, LLMs are you know a use case where some of these things, you know, that's that's part of the thing.

Like, you know, async matters when you are fast API and you're in front of an inference engine.

So if you want to connect, so if I if I've got a worker that I want to make a call to a remote server and I'm gonna have to wait for that server, or it might it'll either stream back to me or I'll have to wait a long time to for it to return its response and then return.

I want async because I want to be able to wait for that whilst I'm handling other requests on, you know, while that's happening.

Yeah.

I mean, we were trying to do a little bit with that with the third party with the ecosystem page, the steering handle, the new steering handle.

Like it's a fantastic page, it's very in-depth.

It's just completely buried, you know, no one can find it.

Um, you know, that's a separate thing.

People I mean, people are working on this, right?

Search and you know, have search, have the search bar on the homepage, which right now we don't.

It's in the docs, which you have to click to.

I was there.

We got some things achieved, but we weren't very able to get very much achieved.

And I think the website really does need a you know well

Let's let's move on.

So we wanted to do um two things.

I think so.

We want to mention a new thing.

We're gonna mention some projects that we're excited about.

Um so Nano Django, Richard Terry gave, I believe, I forgot I missed it.

So I don't know if it was a talk or a lightning talk.

Lightning talk.

Um but this there's a lot of enthusiasm around this, around just a lightweight Django and you know making it not so scary for newcomers.

I need to go see his his talk.

I missed it.

playground on, so I think it's

It's all very cool.

Um and he's doing a great job.

And I think, you know, um, if you've got a big application and you there's an endpoint which you can't you just can't take down while you s while you you know while you update the rest of it, you can spin that endpoint out into a little nano nano Django side point, and that that endpoint can keep serving while you update the rest of the site.

You know, it's not it's not just for the hello world getting getting started examples, but he's doing a really good job pushing that forward and adding the sort of dotting the i's and crossing the t's on these these micro Django examples that we've been playing with for years and years.

Yes.

So my my really exciting one, Adam Hill, um uh DJ Lite, which is um a really simple Django application to you just add it to your installed apps and it tweaks your setting.

There might be one or two out of it, but you it tweaks your settings to give you the right um setting for C SQ Lite in production.

Um there are talks on this and there's blog posts.

Um I s would I put my would I put my you know my business Postgres, I think.

Um and SQLite still doesn't allow concurrent rights, but you can you can teach it to not error and just wait for the other write to finish and then do your right, and that's good enough for almost all you're doing.

Like it's not a wild idea.

probably in there.

But anyway, the put the reason why I like DJ Lite is it wraps you before previous up until this you had to go and search and you had to read the blog posts and you had to it extract exactly what was the right configuration.

Was now you can just pip install DJ Lite, follow the follow the two-step setup, and you're done.

Yeah.

We've already covered too long already.

So the the the it's what it is is a set of really uh rich uh HTML components with the JavaScript all built in and all the rest that give you all the you know just a whole gallery of um form widgets that are more advanced, more elaborate, and you can, you know, and it's it's it's all modern, and you can include just the the JavaScript for the mod for the widgets you need to use and keep your page weights light and all the rest.

You don't have to build in the whole library which is very big.

The intro um post, uh the intro bit of the docs where it talks through the the quick start.

It's it's it's a bit dry and it doesn't really motivate motivate you, but it's a wonderful library.

And what I recommend is skip that bit, come back to that when you've go just go and install it and download a demo and build like one or two of these widgets in in action and see how it easy is.

So give it half an hour doing that rather than reading the docs per se, and then come back to the explanation in the docs once you've convinced yourself actually this is quite cool, because it's a wonderful project, and he's been putting effort into that for years, and it's immensely high quality, and it's I think probably for me the most underrated project in the Django ecosystem.

So anyway,

I got I got to spend some time with him, which was great.

So last thing, um, we're gonna try book recommendation.

So I've been battering, I've been so on my personal side, I've been doing trying to do some book reviews.

Um, and I would I've been talking your ear off about Empire of AI.

So I've been reading all of that.

The Cliff Notes.

She's um Karen Howe.

Sorry, say her name wrong.

Um, has been giving great video interviews, and it really just ties together a lot of the um yeah, the technical, the cultural, everything else about AI in a way that I haven't seen, even though I've read a lot of these books.

So I highly recommend it.

Um you have you have one that's up.

I'm I'm gonna mention uh refactoring to Rust, which is um from Manning.

Uh it's a great one.

The first chapter is why refactor to Rust.

And clearly it's so you can sit with the cool kids at the lunch table, right?

What else?

Like it it sort of implies you could, but you couldn't.

But if you already have dabbled with Rust a little bit, it's quite a good chapter to lock in some of the concepts that you might have come across in other books, but it's not a learn Rust in from that chapter, I promise you.

But it's got chapters on integrating with C, integrating with blah blah blah.

Um, but the Python chapters are really solid.

They show you how to um embed Python in Rust and Rust in Python, and they show you about benchmarking, which is just key, right?

It's like, am I I'm making all these changes, I'm using this tech, am I making anything better?

Actually talks about that and shows how you might, you know, you might you know demonstrate whether you're making things better or not, which is sorely missing in my experience.

So anyway, I'm I I read some of that over the summer and I was like, yes, that's that's a quality refactoring to cost.

Nice.

We're gonna do do our outro again.

Thank you to HackSoft for sponsoring your development partner Beyond Code.

Uh more about their links in the description.

And give us feedback, everyone.

Leave a comment on YouTube, let us know separately.

But we're gonna try doing videos all fall.

We're gonna have some guests, but also go topic driven, and the next week we'll talk about the Django survey and the future of Django.

Okay.

All right.

Bye bye.

Bye bye.